Date: Nov 28, 2012 6:27 PM
Author: Ray Koopman
Subject: Re: Results (!!) on average slopes and means for a1_N_1_C (complement<br> instead of core subset)

On Nov 28, 9:37 am, djh <halitsk...@att.net> wrote:
> Results (!!) on average slopes and means for a1_N_1_C (complement
> instead of core subset)
>
> Len
> Int Avg Slope Mean u'
>
> 1 -2.225882168 0.482402362
> 2 -2.315512399 0.469544417
> 3 -0.769858117 0.485742217
> 4 -1.697049757 0.451420560
> 5 -2.069842267 0.459536902
> 6 -4.427566827 0.457327711
> 7 -0.941379623 0.458781950
> 8 -2.069096413 0.445826306
> 9 -1.620229799 0.442040277
> 10 -3.764328472 0.449422937
> 11 -7.882327621 0.458400090
> 12 -11.82556530 0.458971482
>
> I don?t know if the above results, when compared to the results in the
> previous post for a1_N_1_S, do or don?t indicate that your definition
> of average slope is OK.
>
> As in the a1_N_1_S case, Mean(u?) inversely correlates with LenInt.
>
> But unlike the a1_N_1_S case, average slope ALSO inversely correlates
> with LenInt.
>
> I can readily make a scientific interpretation of these two sets of
> results, but don?t want to do so if you think that these two sets of
> results indicate a problem with the definition of average slope.


The plots look nice, but each point needs standard error bars.
The average slope is a1 + 2*a2*mean_x, so its standard error is
sqrt[ var(a1) + 4*var(a2)*(mean_x)^2 + 4*cov(a1,a2)*mean_x ],
with df = n-3.